410 Chapter 12 Multiple Comparisons Among Treatment Means
12.11 Calculate the Tukey test on the data in the example in Table 11.2, and compare your results
to those you obtained for Exercise 12.10.
12.12 Why might you be more interested in running specific contasts on the data referred to in
Exercises 12.10 and 12.11?
12.13Run Games and Howell (1976) approach to Tukey’s HSD procedure for unequal sample
sizes on the data in Exercise 12.12.
Group 12345
10 18 19 21 29
85879
7.4 8.9 8.6 7.2 9.312.14 Use the Scheffé test on the data in Exercise 12.13 to compare groups 1, 2, and 3 (combined)
with groups 4 and 5 (combined). Then compare group 1 with groups 2, 3, and 4 (com-
bined). (Hint: Go to the discussion at http://www.uvm.edu/~dhowell/methods7/Extras/Unequal-n-
contrasts.html)
12.15 Apply the Tukey procedure to the log transformed THC data from Table 11.6 (p. 339). What
is the maximum FWfor this procedure?
12.16 Apply Dunnett’s test to the log transformed data in Table 11.6.
12.17 How could a statistical package that did not have a Bonferroni command be used to run the
Bonferroni t test on the data in Exercise 12.7?
12.18 The Holm test is referred to as a modified sequentially rejective procedure. Why?
12.19Fit linear and quadratic trend components to the Conti and Musty (1984) log transformed
data in Table 11.6. The control condition received 0 mg of THC. For purposes of this exam-
ple, assume that there were 10 subjects in all groups. (You could add a 2.56 to the 0.5 mg
group and a 2.35 and 2.36 to the 1 mg group without altering the results appreciably.) The
linear coefficients (calculated with unequal spacing on the independent variable) are
[ 2 0.72, 2 0.62, 2 0.22, 0.28, 1.28]. The quadratic coefficients are [0.389, 0.199, 2 0.362,
2 0.612, 0.387].
Verify your answers using SPSS ONEWAY if you have it available.
12.20Calculate the Benjamini-Hochberg test on the data in the example in Table 11.2, and com-
pare your results to those you obtained for Exercise 12.10.
12.21 Use any statistical package to apply the REGWQ (if available), and Scheffé procedures to the
data from Introini-Collison and McGaugh (1986), described in the exercises for Chapter 11
(p. 356). Do these analyses for both Epineq.dat and Epinuneq.dat, which are on the book’s
Web site. Do not combine across the levels of the interval variable.
12.22 In Exercise 12.21 it would not have made much of a difference whether we combined the
data across the three intervals or not. Under what conditions would you expect that it would
make a big difference?
12.23Using the data in Epineq.dat, compute both the linear and quadratic trend tests on the
three drug dosages. Do this separately for each of the three intervals. (Hint: The linear
coefficients are [ 2 0.597110, 2 0.183726, 0.780836], and the quadratic coefficients are
[0.556890, 2 0.795557, 0.238667].)
12.24 Interpret the results in Exercise 12.23.
12.25 Stone, Rudd, Ragozzino, and Gold (1992) investigated the role that glucose plays in mem-
ory. Mice were raised with a 12 hour light-on/light-off cycle, starting at 6:00 AM. During
training mice were placed in the lighted half of an experimental box and given foot shock
when they moved into the dark half. The mice quickly learned to stay in the lighted half.
The day/night cycle was then advanced by 4 hours for all mice, which is known to interfere
with memory of the original training. Three days later mice were retested 30 minutes after
being injected with 0, 1, 10, 100, 250, or 500 mg/kg of sucrose. The purpose was to see
whether sucrose would reduce the disruptive effects of changing the diurnal cycle, andsj^2njXj